While using my message reader (written in JavaScript), I've noticed that if it has a sub-board open with a MessageBase object and lets the user post a message in that sub-board, in Linux, the MessageBase object doesn't seem to get updated with the number of messages in the sub-board. In
Re: Posting a message with messagebase open
By: Psi-Jack to Nightfox on Sat Jul 11 2015 21:19:17
While using my message reader (written in JavaScript), I've noticed
that if it has a sub-board open with a MessageBase object and lets
the user post a message in that sub-board, in Linux, the
MessageBase object doesn't seem to get updated with the number of
messages in the sub-board. In
Yeah.. I've noticed this as well..
Along with, and it's not at all consistent, sometimes when posting a
message it just crashes and hangs up, without ever posting the
message, and I don't think this is a SlyEdit issue either as it
happends just after the save occurs and only then.
I've noticed that sometimes too, in Windows. It seems to happen randomly, so I don't know how to reproduce it. My guess would be that if you happen to be posting a message at the same time that the sub-board messagebase is being updated (via a QWK update), maybe there's some resource conflict that causes a crash. I haven't verified that theory though. I made a local update to my message reader to close its MessageBase object while posting a message and re-open the MessageBase after the posting is done - That seems to allow the total number of messages to be updated (in Linux), and hopefully it will make it more stable too. I'll release an update to my reader after more testing.
I'm currently testing an idea of actually running Synchronet on
CentOS 7, and so far, after grinding thoroughly through building
custom RPMs for DOSEMU, BinkD, and Husky Tools, the main software
needed for my full setup, I am actually coming up with astonishing
results.
That's cool. CentOS is definitely a fairly good & stable server OS (since it's based on RedHat). I'm currently running Mint Linux on my BBS machine (although my BBS is running in a Windows XP VM), but I've considered possibly switching my BBS over to Linux at some point. Mint Linux seems to work well for Synchronet so far, but I had considered CentOS too.
I even went above and beyond and made a DOS door maintenance script
that would parallel run through each door that was configured in the
script to allow running 2+ dosemu sessions at the same time one for
each door, and tracking of each one.
Unfortunately the one limitation was: I ended up using 4DOS 8.00
(Freeware), to get proper logging done. I could never get FreeDOS's
command.com to allow me to capture the date and time into a variable
to output to a logfile, but this allows me now to track that each
door actually ran to completion, and not aborted out in a DOSEMU
crash. :D
I'm not sure I'd consider 4DOS a limitation, since (I believe) it's fully compatible with command.com. It's cool to hear that 4DOS is now freeware. I used to use 4DOS in the early 90s, when DOS was my primary OS, and I thought it provided some useful enhancements over the command.com included with MS-DOS.
Is the file system in use on Linux a shared file system (SMB/CIFS or NFS) or just a local file system?
Is the file system in use on Linux a shared file system (SMB/CIFS
or NFS) or just a local file system?
I can answer that one. Local. :)
I suspect there may need to be a buffer flush (or something comparable) added somewhere. A small test case (e.g. that could just be run with jsexec) would be helpful in isolating and resolving the problem.
Is the file system in use on Linux a shared file system (SMB/CIFS
or NFS) or just a local file system?
It's a local file system.
I just committed an additional file stream flush to the SMB "get status" function (in src/smblib). Please try this and if that doesn't fix it, provide more details or better yet, a small script to reproduce the problem (e.g. using jsexec).
So if you see this reply, that means I didn't crash after posting after a QWKnet packet.
What did happen during that however was:
evnt Finished Importing QWK Network Packet from VERT: (2 msgs) in 1 seconds (2 msgs/sec)
Well, that went through without dropping me.
Next test, inbound FTN messages, as per from sbbsecho.
Results:
Imported: 3 msgs in 0.0 sec (5806.5/min 96.8/sec)
Let's see....
I am using primarily Hyper Allocation for my message bases. Though
I've considered switching to Fast or Self-Packing, not knowing really
what pitfalls each method has..
The message base allocation method shouldn't make any difference with the problem you're describing.
Since no one else is reporting the problem you're seeing, I suggest trying to "undo" some of your mods. Or just as a test, don't use of the 3rd party mods (e.g. message reader) you might be using.
The message base allocation method shouldn't make any difference
with the problem you're describing.
Since no one else is reporting the problem you're seeing, I suggest
trying to "undo" some of your mods. Or just as a test, don't use of
the 3rd party mods (e.g. message reader) you might be using.
I've actually seen what I think he is describing. It seems to happen when I reply to a message using my reader. It happens rarely and randomly, so it's difficult to reproduce. I've noticed it in Windows (I think Psi-Jack is running Synchronet in Linux?).
The other day I made a small JavaScript script that opens a sub-board with a MessageBase object, lets a user post a message while that MessageBase object is open, then closes the MessageBase object. I was trying to reproduce the issue in Linux where if a JS script has a MessageBase open and lets the user post a message in that sub-board, the MessageBase wasn't getting updated as far as the number of messages. I ran that small script on my Linux test setup, and while that script didn't reproduce that issue, I did see it crash Synchronet multiple times after I had saved my message. I tried it in my Windows Synchronet setup, and I didn't see it causing a crash there. This is the code (substitute the value of the subCode variable with a valid sub-board code on your BBS):
load("sbbsdefs.js");
subCode = "loc_general";
var msgbase = new MsgBase(subCode);
if (msgbase.open())
{
console.print("\1n\r\n# of messages (before posting): " + msgbase.total_msgs
+ "\r\n\1p");
var msgHdr = msgbase.get_msg_header(true, msgbase.total_msgs-1, true);
if (msgHdr != null)
{
bbs.post_msg(subCode, WM_NONE, msgHdr);
console.print("\1n\r\n# of messages (after posting): " + msgbase.total_msgs
+ "\r\n\1p");
}
msgbase.close();
}
I'm not sure I'd consider 4DOS a limitation, since (I believe) it's
fully compatible with command.com. It's cool to hear that 4DOS is
now freeware. I used to use 4DOS in the early 90s, when DOS was my
primary OS, and I thought it provided some useful enhancements over
the command.com included with MS-DOS.
I loved 4dos. Such awesome enhancements for command.com's pointlessly lobotomized batch features.
Re: Posting a message with messagebase open
By: Digital Man to Nightfox on Wed Jul 15 2015 16:46:12
I just committed an additional file stream flush to the SMB
"get status" function (in src/smblib). Please try this and if
that doesn't fix it, provide more details or better yet, a
small script to reproduce the problem (e.g. using jsexec).
I hadn't actually commited the flush call yet. I just did. Please
try again.
I updated my Synchronet build on my Linux test system, and that seems
to have fixed the issue I was seeing.
Okay, good. There will probably be other flush-related commits to smblib since this highlights an issue with stale read-data that I was not aware of.
This does not explain any crashes or unintended program exits you were seeing.
I just wonder if the same issue MAY had been causing the disconnects
as well. As you say, it doesn't explain it, but if it resolves it at
the same time...
I can't think of any logical connection between the two things.
I just wonder if the same issue MAY had been causing the
disconnects as well. As you say, it doesn't explain it, but if it
resolves it at the same time...
I can't think of any logical connection between the two things.
Yeah, definitely isn't. A message or two after this reply I made, I
was dropped just while putting in the To, then Subject, then boom, disconnect.
So, Not sure. Definitely still happening, with current CVS codebase.
Does it happen when you don't use the custom message reader mod?
If it's a JS exception causing the termination, there should (normally) be some indication in the log output. I suppose it's possible that the script could handle the exception (whatever it is) and just terminate silently, but that's not normal.
The problem may actually be SyncTERM itself dropping the connection..
A potential situation I pondered.
The way I'm currently testing this theory is sadly by using NetRunner.
I hate this particular terminal app because it's so slow and clunky,
and doesn't even support rlogin or ssh like I usually use. heh. And
it's slow, even over LAN.
So far, though, using NetRunner, I have yet to see a single
disconnect, but I'm going to continue using it for the rest of this
week to see.
And you could try different protocols in SyncTERM (e.g. Telnet instead of SSH).
So far, though, using NetRunner, I have yet to see a single
disconnect, but I'm going to continue using it for the rest of this
week to see.
And you could try different protocols in SyncTERM (e.g. Telnet instead
of SSH).
That reminds me, several years ago I noticed that when I was connected to my BBS via SSH (using SyncTerm), I would get disconnected from my BBS after some amount of time. It didn't seem to be related to posting messages or anything in particular. But I noticed that seemed to be fixed in more recent builds of Synchronet/SyncTerm.
Re: Posting a message with messagebase open
By: Psi-Jack to Nightfox on Sat Jul 11 2015 21:19:17
This is another message from you that has somehow migrated to another board - I saw this in the Sports board on Dove-Net. This was originally in the Synchronet JavaScript board.
Sysop: | Chris Crash |
---|---|
Location: | Huntington Beach, CA. |
Users: | 585 |
Nodes: | 8 (0 / 8) |
Uptime: | 26:44:08 |
Calls: | 10,757 |
Files: | 5 |
Messages: | 452,124 |